[Boost-users] Re: A boost.graph edge problem

From: Jeffrey Holle (jeff.holle_at_[hidden])
Date: 2004-09-22 21:14:12

Found my problem!
It was my ltSubgraph functor.
Instead of:
   return s1.m_source < s2.m_source ||
       (!(s1.m_source < s2.m_source) && s1.m_target < s2.m_target);
It needed to be:
   return s1.m_source < s2.m_source ||
       (!(s2.m_source < s1.m_source) && s1.m_target < s2.m_target);

Jeffrey Holle wrote:
> I'm using boost 1.31.0 and gcc 3.4.2.
> 
> While this problem is more of an application code problem, since I got 
> no help on other lists, I trying this one with hopes of a response.
> 
> I need to identify parallel edges from my graph, which is defined as:
>  typedef subgraph<adjacency_list<vecS,
>                                  vecS,
>                                  bidirectionalS,
>    property<vertex_index_t,int,VertexProperty>,
>                                 EdgeProperty> > DataGraphT;
> typedef boost::graph_traits <DataGraphT>::edge_descriptor DataEdge;
> 
> In the test case that I'm working on, I know I have parallel edges, 
> actually 4 sets of 3.
> 
> I'd like to use this code:
> DataGraph::handleParallelEdges(void)
> {
>   typedef multiset<DataEdge,ltSubgraph> Edges;
>   Edges edges;
>   graph_traits<DataGraphT>::edge_iterator ei,ei_end;
>   for (tie(ei,ei_end)=boost::edges(dataGraph_);ei!=ei_end;++ei)
>       edges.insert(*ei);
>   for (Edges::const_iterator iter = edges.begin(); iter != edges.end();
>     advance(iter,edges.count(*iter)))
>     if (edges.count(*iter)>1)
>       cout << "See duplicate" << endl;
> }
> 
> However, when count never returns.
> The likely cause of this (I suspect) is the ltSubgraph functor.
> Its definition is:
> 
> struct ltSubgraph : public std::binary_function<DataEdge,DataEdge,bool>
> {
>   bool operator() (const DataEdge& s1, const DataEdge& s2) const
>   {
>     return s1.m_source < s2.m_source ||
>      (!(s1.m_source < s2.m_source) && s1.m_target < s2.m_target);
>   }
> };
> 
> I say this because I can see that this functor is being continuously 
> called in this infinite loop.
> 
> Note that I use this functor in multiple other cases, but not with an 
> associate container that allows duplicate keys.
> 
> Is there a problem with this code?